Dementia Hospital Discharges (Alzheimer's Disease and Related Dementias)

Limitations

Inpatient stays of Georgia residents in Georgia's non-federal hospitals whose principal diagnosis is in OASIS's ADRD list (dementias plus Parkinson's, Huntington's and CJD), by home county: counted per stay, not per person; stays in other states are missing. all is OASIS's total and the 8 groups add up to it, but subtypes nest and overlap: never add causes. OASIS withholds rates on 1-4 stays. Stays of unknown ethnicity count only in race_unknown, so race-group rates run low (NULL above 20%). Age-adjusted rates are OASIS's: weighted averages of age-specific rates with the year 2000 U.S. standard million as weights, so places and groups with different age mixes compare fairly; they never add or average across rows. OASIS publishes every count, however small, and withholds rates based on 1-4 events (NULL here); the platform adds no suppression of its own and never recomputes a withheld rate. A group with no residents in a place (num_population 0) has no rate (NULL): OASIS shows a crude rate of 0 there (and withholds the age-adjusted rate itself), but a rate on no population is undefined and would rank as the lowest (our mask; the count stays 0). 70% of rows with residents and at least one stay have no age-adjusted rate, every one of them on 1-4 stays (small counties, small groups and rare causes); every row with residents and 5 or more stays has both rates (race_unknown never has a rate). On 80,662 rows with no stays OASIS also codes the age-adjusted rate -2 (NULL), because one of the group's age groups has no residents in the place (mostly small race groups in small counties): the crude rate of 0 stands beside it. Principal diagnosis (OASIS's page; its alzheimers_disease is the same stays as hospital_discharges_by_cause's in every place, year and group): a stay for another condition in a patient with dementia is not counted, and every stay here is also a stay in hospital_discharges_by_cause, so never add the two topics. The groups never overlap (one principal diagnosis per stay). Within them, the subtypes of alzheimers_disease and dementia_in_other_diseases add up exactly, but those of other_degenerative_nervous_system_diseases add to more than the group (Georgia 262 stays in 2024, its subtypes 370): neurocognitive_disorder_with_lewy_bodies and corticobasal_degeneration (103 and 6 in 2024) are also counted in other_specified_degenerative_nervous_system_diseases (146), as OASIS's own selected total, which counts each stay once, confirms; counted once, the subtypes fall 1 to 6 stays a year short of the group in Georgia. A discharge is one inpatient stay in a non-federal acute-care hospital in Georgia, counted in the year the patient left (alive or dead), by the patient's county of residence: a person admitted twice counts twice, newborns' birth stays and deliveries are included, and stays of Georgia residents in other states' hospitals or in federal (VA and military) hospitals are not, so counties near another state's hospitals may look low. OASIS's hospital data records ethnicity from 2016, this module's first year (Georgia: 28 Hispanic stays in 2016, 112 in 2025), but some hospitals still record none: race_unknown is 11.2% of Georgia's stays in 2016 and 0.2% to 5.5% a year since, up to 68% of a county's in a year (counties with 50 or more stays); OASIS advises caution with Hispanic figures (https://oasis.state.ga.us/dtt/content/KDIADRDMorb.html). Where race or ethnicity is unknown for more than 20% of a place-year's stays (race_unknown / all), the race groups' counts and rates for that place and year are NULL (our mask, OASIS's own 20% threshold for missing items): each group's count would miss its share of the unknowns, so its rate runs low and its trend follows the recording, not the group. The all, sex and race_unknown rows are kept, so the unknown share stays visible. It applies to 78 county-years (55 counties, never Georgia). race_unknown is our calculation, all minus the seven race groups (unknown ethnicity, or non-Hispanic of unknown race): a count with no population or rate. dementia_with_mood_disturbance and dementia_with_psychotic_disturbance are ICD-10-CM codes added in October 2022, so no stay could carry them before: they have no rows before 2022 (rows dropped, not served as zeros; 2022 covers October-December only), and no stay before 2024 (psychotic: 2 stays in Georgia that year) or 2025 (mood: 1). Causes are coded in ICD-10-CM throughout (the October 2015 switch from ICD-9-CM is before the module's first year). Coding moved stays between causes: Georgia's vascular_dementia fell from 567 stays in 2021 and 437 in 2022 (the last quarter of 2022 already used the new codes, our reading) to 85 in 2023, while unspecified_dementia rose from 1,026 to 1,215 and 1,551, after ICD-10-CM expanded its dementia codes in October 2022 (our reading), and vascular_dementia was back to 294 in 2025: read the two together, or all, from 2022 on. The Alzheimer's subtypes also shift with coding while alzheimers_disease holds steady (1,204 stays in 2024, 1,201 in 2025): other_alzheimers_disease fell from 197 to 60 as alzheimers_disease_unspecified rose from 838 to 954, and alzheimers_disease_late_onset ran from 134 (2016) to 356 (2019): follow alzheimers_disease, not its subtypes, over time. parkinsons_disease_only nearly doubled (376 stays in 2016, 738 in 2025), as in OASIS's Discharge module, and is part of all, so all's rise (3,839 stays in 2016, 4,484 in 2025) is partly Parkinson's. num_population is OASIS's own population (its population module) for the place, year and group, the denominator of every crude rate (num_discharges / num_population x 100,000 reproduces each published crude rate); the race groups' populations add to all, as do male and female. 2025, the newest year (4,057 stays in 2024, 4,484 in 2025), carries the same alzheimers_disease stays as OASIS's Discharge module, which added it on 2026-09-03; OASIS marks no year provisional. Counties add up to Georgia, except race groups, whose counts are NULL in the county-years the race mask covers (Georgia's race rows count every county, so a race group's counties add up to less there). Not served: OASIS's SES-vulnerability rows and payer 'rates' (each divides by the whole place's population; the payer rows stay in bronze) and its age breakdown. tract_geoid is always NULL.